Filtration Is a Necessary Condition for Discovery and Persistence

This paper introduces Filtration Theory as a structural account of persistence and discovery in complex systems. The core claim is that systems capable of sustained exploration, learning, and adaptation must enforce invariant-first control: non-negotiable constraints applied prior to capability expansion, optimisation, or learning. Rather than treating filtration as a secondary safety mechanism or heuristic design choice, the paper argues that selective boundary enforcement and disorder export are necessary structural conditions for persistence across domains. Systems that fail to constrain admissibility before expansion exhibit instability, collapse, or unbounded drift. The framework is domain-general and applies to artificial intelligence, biological regulation, organisational governance, and scientific discovery processes. The paper formalises the distinction between participation and direction, and clarifies why discovery requires controlled openness rather than unconstrained exploration. The argument is explicitly falsifiable and is supported by a canonical architectural diagram archived separately on Zenodo. This work is intended as a preprint and conceptual foundation for further empirical and formal investigation.